File Name : esi fig. 1.eps Caption : esi fig. 1. comparison of variation in peptide intensity between printing conditions. percent coefficient of variation (cv) of 1x, 2x, and 1x2 assays are displayed for corresponding inter-assay peptide intensities from affinity screening. peptides within the red box represent the top 5% of peptides, identified as high affinity polysialic acid (psa)-binding peptides (11 peptides total; table 1 and esi table 1). intensities are shown in relative intensity units. File Name : esi fig. 2.eps Caption : esi fig. 2. peptide residue compositional analysis. residue occurrence in the partial peptide library (consisting of all mab epitope mapping-derived peptides and unmodified phage display screening-derived peptides) is compared to occurrence in the top 10% of this peptide set, where top 10% refers to peptides with highest affinity binding in the partial library (14 peptides). statistical significance of an increase or decrease in occurrence of the following residues could not be determined as low residue occurrence in the sample population precluded the assumption of normal distribution: d, e, h, c, w, a, i, and m. in contrast, decrease in t and l and increase in y and v were deemed not significant. (two-tailed z test for population proportions; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). the greatest relative change in occurrence is evident for basic amino acids r and k (positively charged at the screening ph of 7.4), followed closely by n. it is possible that the changes seen for s and g are influenced by biases from phage display screening, as well as simultaneous evaluation of phage-based and other peptides with the microarray platform. specifically, the high initial frequencies of g and s arise from c-terminal linkers included to mimic phage display conditions, and decreases in these frequencies are expected if high binders are not assumed to arise solely from phage display-derived peptides. File Name : esi fig. 3.eps Caption : esi fig. 3. selectivity of high affinity peptides (table 1) with (a) 2x and (b) 1x2 printing (intra-assay average intensities from triplicate measurements shown in relative intensity units). blue circles = 10 μm colominic acid (ca) with 0 μm chondroitin sulfate a (cs-a), green circles = 10 μm ca with 0.1 μm cs-a, orange circles = 10 μm ca with 1 μm cs-a, and red circles = 10 μm ca with 10 μm cs-a. as with the 1x condition (fig. 3), peptides of similar binding intensities in affinity screening display different levels of binding ability in the presence of 10% competitor, with peptides 79 and 214 demonstrating relatively higher selectivities. no high affinity peptides show significant (i.e., above background) levels of binding to psa with equimolar concentrations of cs-a. it is possible that the generally lower selectivity displayed by these 11 high affinity peptides in the 2x and 1x2 conditions as compared to 1x (fig. 3) are due to higher peptide densities (ligand density has been shown to affect both affinity and selectivity of carbohydrate interactions¹). File Name : esi fig. 4.eps Caption : esi fig. 4. selectivity of peptides derived from peptide 106 (p106; phage display screening peptide) with (a) 2x and (b) 1x2 printing (intra-assay average intensities from triplicate measurements shown in relative intensity units). blue circles = 10 μm ca with 0 μm cs-a, green circles = 10 μm ca with 0.1 μm cs-a, orange circles = 10 μm ca with 1 μm cs-a, and red circles = 10 μm ca with 10 μm cs-a. p106 series indices corresponding to primary peptide library indices are shown in table 3. peptide 29 in the above plots (library index 170) displays high affinity as in the 1x assay (fig. 4), likely due to the replacement of neutral residues with three arginines (among other substitutions). however, this peptide shows poorer selectivity in the presence of 10% competitor than peptides of similar affinity, suggesting that designing selective peptide ligands to target a negatively charged polysaccharide requires considerations beyond a general increase in peptide basicity (such as positioning of charged residues, peptide conformation, etc.). File Name : esi fig. 5.eps Caption : esi fig. 5. psa-binding of mab735-derived peptides. average binding intensities from affinity screening in 1x, 2x, and 1x2 assays (with triplicate intra-assay readings) are displayed in relative intensity units. though differences in assay conditions (specifically peptide density) may influence variation seen in peptide affinity to psa, inter-assay intensities are plotted to display general trends in binding. error bars represent standard deviations of inter-assay values. similar binding of contiguous sequences is in keeping with derivation of sequences from epitope mapping, with adjacent sequences containing 13 residue overlaps. higher binding regions do not necessarily correspond to cdrs of mab735 (cdrs from which sequences originate shown above corresponding points), as mab735 has a discontinuous epitope in the binding of oligosialic acid.² however, linear epitope mapping of the discontinuous epitope may have provided peptide fragments encompassing residue composition and positioning with a higher likelihood of target interaction,³ similar to a rational partners-based design approach in the engineering of peptide ligands to protein targets.^4–6 additionally, future work with linear epitope mapping of sialic acid-binding proteins with continuous binding pockets may help elucidate lectin–psa interactions, an advantage not afforded by random screening. File Name : esi fig. 6.eps Caption : esi fig. 6. tbe page for molecular weight assessment of glycans. 4–20% novex^tm tbe gel (invitrogen, carlsbad, ca) with lanes 1–2 ca (sodium salt, nacalai usa, inc.) and lanes 3–4 cs-a (from bovine trachea, sigma-aldrich); 210 μg loaded in lanes 1 and 3 and 52.5 μg loaded in lanes 2 and 4. gel stained with alcian blue (image acquisition in grayscale). given polydispersity of samples, rough molecular weight of cs-a estimated as 45 kd based on the following: i) manufacturer provided average molecular weight of ca = 30 kd and observed molecular weight range of ca, ii) expected migration of oligonucleotide fragments and proteins of different molecular weights (invitrogen gel migration chart for novex^tm pre-cast gels),^7,8 and iii) estimates of bovine trachea cs-a molecular weight from literature.^9,10 each of the previous likely provides a very broad estimation of molecular weight, and significantly more accurate determination may be performed chromatographically. however, for the purposes of this work, a general estimate suffices for selectivity studies, where selectivity conditions chosen (0, 0.1, 10, and 100% competitor with ca) represent a broad range of low to high (with logarithmic intervals), and relative amounts of competitor between conditions are accurate. additionally, conclusions on selective binding of peptides are relative (between assay or selectivity conditions or between peptides) and are not based on constants (such as k_d) dependent on glycan concentrations. the term “equimolar,” when used to describe cs-a and ca amounts, is thus meant to indicate a general equivalency in amounts of glycans that encompasses error from molecular weight estimation, including from sample polydispersity. File Name : esi fig. 7.eps Caption : esi fig. 7. spr spectroscopy of positive and negative control peptides (see esi table 3) with the biacore 3000 system (ge healthcare, chicago, il). peptides were synthesized at >95% purity (confirmed with hplc) by biomatik with n-terminally acetylation and c-terminal cysteine addition. peptides were immobilized by thiol coupling of cysteines to cm5 sensor ships in accordance with standard protocols (ge healthcare). ca solutions in pbs, ph 7.4 from 0–1.67 mm (0–50 mg/ml) were injected for 180 seconds at a flow rate of 10 μl/min, and the surface was regenerated with 2 m nacl injection for 90 seconds at the same flow rate. sensorgrams were obtained following subtraction of reference flow cell response. equilibrium binding responses at each analyte concentration were determined as the relative response difference from baseline at which binding reached saturation, and equilibrium binding curves were subsequently generated in matlab using the four-parameter logistic binding model.¹² a k_d value of 575±74 μm was thus obtained (error represents standard error of the fit with 95% confidence bounds). given the unknown absolute ligand density (as for microarrays) and the effect of ligand immobilization on binding constants, though a stoichiometry of 1.32±0.15 was obtained from the fit, it is possible that the affinity constant is an overestimation of 1:1 in-solution binding strength. additionally, the high viscosity of ca contributed to artifacts within sensorgrams (such as spiking) due to high refractive index changes with analyte injection; extraction of binding values was possible with artifact removal, but such effects may have contributed to error. (artifacts were minimized with immobilization of ca and injection of peptide; however, responses fell below the detection limit due to the relatively small size of peptides compared to ca.) nevertheless, spr responses confirm the ability of the peptide microarray assay to discriminate between binding and non-binding peptides to the polysaccharide target and furthermore, demonstrate the high-to-low throughput approach that may be pursued for future analysis of psa-binding peptides.